laws and cause

Laws and cause*

Donald DAYIDSON..

Abstract Anomalous Monism is the view that mental entities (objects and events) are identical with

physical entities, but that the vocabulary used to describe, predict and explain mental events is neither definitionally nor nomologically reducible to the vocabulary of physics. The argument for Anomalous Monism rests in part on the claim that every true singular causal statement relating two events is backed by a law that covers those events when those events are appropri- ately described. This paper attempts to clarify and defend this claim by tracing out some conceptual relations among the concepts of event, law, and object.

In her inaugural lecture as Professor of Philosophy in the University of Cambridge, G.E.M. Anscombe examined the “often declared or evidently assumed” view that

causality is some kind of necessary connection, or alternatively, that being caused is - non-trivially - instancing some exceptionless generalization saying that such an event always follows such antecedents.

She complained that “the truth of this conception is hardly debated”, and surveyed its history from Aristotle to the then present to make her point. I have the honor of bringing up the rear: “even Davidson”, she remarks in her last paragraph,

will say, without offering any reason at all for saying it, that a singular causal statement implies that there is such a true universal proposition.

I offer this paper as part of the much deserved celebration in honor of Henri Lauener, valued contributor, and formidable champion of, international philosophy, thoughtful and witty companion, steadfast and affectionate friend.

I wish to thank David Albert and, Noa Latham for their help. The central idea in this paper has a Kantian ring, and was used by Gordon Brittan in his Kant’s Theory of Science, Princeton Universitv Press. 1978. As he Eenerouslv notes, he first heard the idea in mv classes at Stand- ford U$iersity. ‘

I

The Universitv of California at Berkelev, USA Cuusuiity an8 Determination, Cambridge University Press, Cambridge, 1971, p. 1. Ibid., p. 29.

Dialectica Vol. 49, No 2-4 (1995)

264 Donald Davidson

In the paper to which Anscombe refers,3 I offered no reason for saying true singular causal statements (like “The eruption of Vesuvius in 79 A.D. caused the destruction of Pompeii”) imply the existence of laws that cover the case, nor did I offer any in my subsequent article “Mental Events”. I wrote there that I was treating this relation between the concepts of law and cause as an assumption, observing that even someone who was dubious of the assumption might be interested in an argument claiming to show that the assumption, along with other commonly held (though debated) premisses, implied a form of monism - what I called “Anomalous Monism”.

Various critics have joined me in noting that if the assumed relation between laws and causality (which in “Mental Events” I called, rather windily, “The Principle of the Nomological Character of Causality”) is false, the argument for anomalous monism fails, and some have enjoined me to produce a reason to accept the assumption. Others, like Anscombe, have expressed doubt that the assumption is true. “I have to doubt Davidson’s. . . premise, that all pairs of events related by ‘cause’ are subsumed under laws”, writes Jennifer Hornsby.s Ernest Sosa asks, “Why must there always be a law to cover any causal relation linking events x and y? What enables us to assume such a general truth?”6 Qler Burge is more skeptical still: “I do not think it a priori true, or even clearly a heuristic principle of science or reason, that causal relations must be backed by any particular kind of law.”

Burge is right that if there is a reason for holding the cause-law thesis, the argument must in some sense be a priori, for the thesis clearly is not a pronouncement of ordinary logic, nor can it be established empirically, a negative point on which Hume and Kant were agreed. To say the cause-law thesis is a priori is not, of course, to say that particular causal laws are a priori. If the thesis is true, what we know in advance of evidence is that if a singular causal claim is true, there is a law that backs it, and we can know this without know- ing what the law is.

“Causal Relations”, The Journal of Philosophy, 64(1967): 691-703. Reprinted in Es-

‘ First published in Experience and Theory, L. Forster and J.W. Swanson, eds., The

“Which Mental Events are Physical Events?”, Proceedings of the Aristotelian Society,

Ernest Sosa, “Mind-Body Interaction and Supervenient Causation”, in Midwest

’ “Philosophy of Language and Mind: 1950-1990”, in The Philosophical Review, 101

says on Actions and Events, Oxford University Press, 1980.

University of Massachusetts Press, 1970; reprinted in Essays on Actions and Events.

1980-81, p. 86.

Studies in Philosophy, IX, 1984, p. 278.

(1992), p. 35.

Laws and cause 265

I.

The cause-law thesis needs to be made more definite. By a singular causal statement I mean a statement that contains two singular terms (names or definite descriptions) referring, or purporting to refer, to events, joined by some form of the verb “to cause” (if the statement is expressed in Enghsh). Of course other verbs can do the same work, for example “produce”, “result in”, “have as consequence”, etc. Examples of singular causal statements are: “His lighting the match caused the explosion”, “The next California earthquake will cause the destruction of the Golden Gate Bridge”, “The hurricane is the cause of the rise in the water level”. “The hurricane caused the water level to rise” is not, however, a singular causal statement, since “the water level to rise” is not a singular term; this last sentence says only that the hurricane (a particular event) caused at least one event that was a rising of the water level. “The rise” purports to pick out a particular event; “a rise” marks a general existential claim without implying singularity.

Singular causal statements are extensional: their truth value is invariant under the substitution of one name or description of an event for another name or description of the same event. Thus if Socrates was Xanthippe’s husband, and Socrates’ drinking the hemlock resulted in Socrates’ death, it follows that Socrates’ drinking the hemlock resulted in the death of Xanthippe’s husband. The point may seem obvious, and indeed it is; yet it has escaped all those who have been tempted to think that if singular causal statements imply the existence of a covering law, they must imply, or somehow indicate, some particular law that covers the case. It is easy to see why this does not follow. Given the endless possibilities for redescribing events (or anythmg else) in non-equivalent terms, it is clear that there may be no clue to the character of an appropriate law in the concepts used on some occasion to characterize an event. What may be the case is that if a singular causal statement is to be explanatory in some desired sense, it must put its hearer in mind of at least the general nature of a relevant law. It also may (or may not) be the case that the only, or best, reason for believing a singular causal statement is evidence for the truth of some law that covers the case. But such epistemological and explanatory issues, however we resolve them, ought not lead us to color singular causal statements intensional.

In formulating the cause-law thesis, what should we count as a law? Laws must be true universally quantified statements. They also must be lawlike: they must support counterfactuals, and be confirmed by their instances (these conditions are not independent). To quallfv as strictly lawlike, they should contain no singular terms referring to particular objects, locations or times

266 Donald Davidson

(strictly lawlike statements are symmetric with respect to time and location). Strictly lawlike statements do not contain open-ended phrases like “other things being equal”, or “under normal conditions”. It must be admitted that such phrases are, tacitly or explicitly, part of the content of many legitimate laws; thus many laws are not strict, including the laws peculiar to such sciences as geology, biology, economics, sociology and psychology.

The distinction between strict and non-strict laws is essential to the argument for Anomalous Monism. The argument for Anomalous Monism has three basic premises: (1) there are causal relations between events described as physical and events described as mental, (2) there are no strict laws relating events under physical descriptions with events under mental descriptions, and (3) if two events are related as cause and effect, there is a strict law covering the case. The first premise I took to be evident: events in the world we describe in physical terms cause and are caused by thoughts. The second premise I have defended at some length, mainly on the grounds of the uneliminably normative or rational aspect of intentional idioms, and the consequent irredu- cibility of mental concepts to concepts amenable to inclusion in a closed system of laws. The argument went this way: It is plausible that there is a set of concepts (perhaps there are many such sets) which lend themselves to the formulation of a closed causal system. Let us call these concepts the concepts of physics. In this case, for any two events related as cause and effect, there will be a strict law, i.e., a physical law, covering the case. Since mental concepts are not amenable to inclusion in a closed system, the strict laws covering singular causal relations expressed in (at least partly) mental terms must also be expressible in physical terms. Hence events described in mental terms must also be expressible in physical terms: in ontic language, mental events are identical with physical events. It follows that mental concepts are supervenient on physical concepts, in this sense: if two events fail to share a mental property, they wiU fail to share at least one physical property.

Anscombe, as we have seen, attacked the view that causes “necessitate” their effects, and she explained this as requiring a law that is an “exceptionless generalization saying such an event always follows such antecedents”. The notion of necessity comes in, I suppose, with the idea that one can deduce a statement of the existence of the effect from a statement of the cause and the appropriate law. Anscombe takes this to forbid the indeterministic laws of quantum physics. The constraints I have put on the laws that the cause-law thesis says exist do not, however, disallow probabilistic laws. Such laws are universal and are exceptionless (the probabilities they predict have no exceptions). So it is possible, though unlikely, that Anscornbe is not questioning the cause-law thesis as I have stated it. (We do disagree on a consequence of

Laws and cause 267

allowing probabilistic laws, since she holds that such indeterminism leaves room for a meaningful concept of human freedom, while I think the indeterminism of quantum physics cannot facilitate, though it might conceivably sometimes frustrate, freedom of action).

Since it allows probabilistic laws, the cause-laws thesis does not (in one fairly standard sense of that messy concept) imply determinism. Neither, then, does it imply complete predictability, even in principle, nor retrodicta- bility.

II. In the Enquiry Concerning Human Understanding, Hume defines a cause

to be “an object, followed by another, and where all the objects similar to the first are followed by objects similar to the second”. If we think of this as a stipulative definition, then one condition for the existence of a law, generality, is built into Hume’s definition. But if the definition is intended as an analysis of the common conception of cause we can, with Anscornbe, ask if it is correct. It is in any case clear that his definition will not satisfy the cause- law thesis, since it fails to distinguish true but non-lawlike generalizations of the night-day sort from lawlike generalizations. Not that nothing Hume says invites us to observe this distinction; his discussion of induction requires it. Hume’s definition of cause just quoted says that the truth of a singular causal statement depends on the existence of a true generalization that covers not only the case at hand, but all other cases, observed and unobserved, past, present and future. This raises the obvious question what justifies us in believing in the truth of such a generalization, and therefore in the truth of any singular causal statement. Hume’s answer, as we know, is that nothing justifies us in either belief; but then we are left to wonder why we have any such beliefs. For this Hume does have an answer, the gist of which is given by his alternative “definition” of cause: a cause is “an object, followed by another, and whose appearance always conveys the thought of that other”. Here, as the “always” makes clear, “an object” stands for a class of objects. Although this pronouncement has invited many interpretations, I will take it here to imply at least this: we believe that one event has caused another if every event that seems similar to the first iias been followed by an event that seems similar to the second. It adds something to say (as Hume does) that this pairing of sets of events or appearances is a habit of the mind, for this implies a disposition present even when not at work, a disposition to project the pairing be- yond what has been given. We expect fire to burn and bread to nourish in the future.

268 Donald Davidson

Nelson Goodman, in his basic book Fact, Fiction and Forecast8, praised Hume for realizing that only human tendencies to class@ and associate in one way rather than another could be called on to characterize inductive reason- ing, but he criticized Hume on two grounds. The first was for thinking this approach is skeptical; in Goodman’s opinion induction does not need justification, only a correct, naturalistic description. The second was that even taken as description, Hume missed the difficulty that lay hidden in the concept of similarity. Goodman brought this out by inventing the predicate “grue”. Something is grue if examined before some future time t and found to be green, and otherwise is blue. We have no trouble understanding this predicate; it is defined in a straightforward way in intelligible terms. We can therefore recognize things that are grue as similar: the predicate “grue” is true of all of them; they are alike in being grue! Yet we realize that “All emeralds are green” is, if not true, at least lawlike, while “All emeralds are grue” cannot also be lawlike if induction is to provide any guidance to the future. Even if we abandon the search for an ultimate justification of our inductive practices, it is legitimate to ask what those practices, at their reflective best, are. The concept of similarity cannot, by itself, cany the burden of distinguishing the lawlike from the non-lawlike.

Is it quite right, though, to say Hume let matters rest with the unquestioned concept of similarity? According to Goodman,

The real inadequacy of Hume’s account lay not in his descriptive approach but in the imprecision of his description. Regularities in experience, according to him, give rise to habits of expectation; and thus it is predictions conforming to past regularities that are normal or valid. But Hume overlooks the fact that some regularities do and some do not establish such habits; that predictions based on some regularities are valid while predictions based on other regularities are not. . . Regularity in greenness confirms the predic- tion of further cases; regularity in grueness does not. . . Regularities are where you find them, and you can find them anywhere.

If we take Hume’s first definition of cause, this criticism is apt. But Hume’s reformulation defines cause in terms of the inductions we actually make, not those we might have made had we been differently constituted. When Hume writes that “after a repetition of similar instances, the mind is carried by habit, upon the appearance of one event, to expect its usual attendant, and to believe that it will exist” lo we notice that “similar” and “usual” are uncritically

First published in 1979. Page numbers here are from the fourth edition, Harvard

bid, p. 82. An Enquiry concerning Human Understanding, Sect. VIII, Part 11.

University Press, 1983.

Laws and cause 269

used. But the second definition of cause doesn’t employ these words, and may be viewed instead as defining the relevant concept of similarity. To take one of Hume’s examples: if every time we have observed a certain vibration of a string it has been followed by a certain sound, then the next time we observe that vibration, we expect that sound and not another. The phrases “a certain vibration” and a certain sound” assume the classifications that are appropriate to inductions; but we may take the operative expectations to fix the classifications. Our expectations or “projections” thus distinguish the lawlike from the non-lawlike. It is true that this account differs from Goodman’s in several ways, but not, as he suggests, in that Hume provides no answer at all to Good- man’s “New Riddle of Induction”.

Goodman’s analysis is, of course, far more explicit, detailed and precise than Hume’s. It also differs in another interesting respect. Goodman’s detailed account makes the lawlike status of a statement depend on the projectible status of its individual predictions; Hume’s account rather treats whole statements (sentences) as lawlike or not (which depends on present habits as formed by past experience). Thus Hume’s analysis allows that a generalization of the form “All Fs are G’s” may be lawlike and “All F’s are H’s” not, even though Hmay occur in other statements that are lawlike. In other words, with respect to things that are F, G may be a projectible predicate, while with respect to things that are F, H may not be projectible. Goodman’s analysis does not allow such cases, since for him projectibility is a property of predicates, not of predicaies relative to other predicates. In this respect I think Hume’s line superior, for reasons I shall presently discuss. l1

111.

Hume says (though not with complete consistency) that we believe in a causal connection between two events only when we have experienced repeated conjunctions and no exceptions (I’m not bothering with the further conditions of succession and contiguity here). It is therefore worth pointing out that the cause-law thesis is not committed to this idea. Nor is it committed to the view that the only way of supporting or confirming a singular causal statement is by reference to relevantly similar cases. I am with those philoso- phers (for example Anscombe, Ducasse and McDowell) who think Hume was wrong in supposing we never directly perceive that one event has caused another, even when we have no supporting evidence drawn from similar cases.

l1 I made this point, perhaps too crudely, in “Mental Events”.

270 Donald Davidson

John McDowell recognizes that I do not think causal laws are merely true generalizations (though as I say above, neither do I think this is all there is to Hume’s view), but he argues that if we give up this reduction, it is hard to see why we should accept the cause-law thesis, which McDowell considers part of the “broadly Humean picture of causation”. In other words, McDowell holds that if we give up Hume’s epistemic claim that the only evidence we have for causal connections is observed regularities, “it is hard to see what now holds the [broadly Humean] picture of causation in place.”

Hume’s own recommendation of it is, in effect, that since singular causal relations are not given in experience, there is nothing for causation to consist in but a suitable kind of generality. And this recommendation seems inextricably bound up with a ‘dualism of scheme and content, of organizing system and something waiting to be or- ganized‘, the untenability of which Davidson has done as much as anyone to bring home to us. Without that dualism, there is no evident attraction left in the thought that singular causal relations are not given in experience. l2

The connection with scheme-content dualism is obscure to me, but it is any case irrelevant, since I have never claimed that singular causal relations are not given in experience. The notion of being “given in experience” is not one for which I have felt much need, in this context or any other. But if it means here no more than that the excitation of our senses may sometimes cause us correctly, justifiably and without inference to believe that a particular’event has caused another, then I certainly accept that singular causal relations are often given in experience. l3 McDowell apparently holds that once we grant that it is possible to perceive that one particular event has caused another there is no reason to accept the cause-law thesis. But why should we assume this particular connection, or non-connection, between an epistemic and a non-epistemic issue?

C.J. Ducasse believed that a true singular causal statement entails the existence of a general law, and that it is possible to observe that one event has caused another without having any independent reason to accept a generalization. He held that all knowledge of causality depends on observing particular cases, though if we knew that an event c caused an event e in a situation S, we would then know that any event exactly like cwould cause an event exactly like e in a situation exactly like S. Unfortunately, his argument for this view seems to rest on two confusions that he elsewhere warns against: the confu-

John McDowell, ‘‘Functionalism and Anomalous Monism”, in Actions and Events: PeRpCctiws on the Philosophy of Donald Davidron, E. -re and B. McLaughlin, eds., Blackwell, Oxford, 1985, p. 398. The inner quote from Davidson is from Znquiries into Tiuth and Interptetation, Oxford University Press, 1984, p. 189. * This has always been my view. See “Actions, Reasons and Causes” in Essays on Ac- tions and Events, p. 16.

Laws and cause 271

sion of particular events with events of “the same” type, and the idea that the concept of a sufficient condition can be applied to events as well as to sentences about, and descriptions of, events. l4 What I find of interest for my present purpose is his definition of cause.

Ducasse tells us that he performed the following “experiment” with his students: He would put a paper-covered cardboard box on his desk, and ask the students to keep their eyes on it. He would then place his hand on the box, and the end of the parcel facing the students would instantly glow. Ducasse next asked them what caused the box to light up when it did, and they would all naturally answer that the glowing was caused by Ducasse’s placing his hand on the box. l5 Ducasse notes that the experiment was not repeated, no similar cases were offered for observation. He allows that the students might have the cause wrong; his point was only to establish something about their criteria for judging what had caused the package to glow. On the basis of a single “experiment”, they believed there was a causal connection, as, of course, there was. This leads Ducasse to the following definition of cause: if cis the only change in a situation S which precedes the only subsequent change e in S, then cis the cause of e. This formulation is mine, and it does not do full justice to Ducasse’s more guarded definition. My formulation does, however, bring out a difficulty Ducasse understandably overlooked. He did not pause to ask what constitutes a change, and therefore what sorts of entities could count as causes and effects.

A natural first stab at saying what a change is goes something like this: some predicate Pis true of an object or situation at a given time, t, and subsequent to t Pis no longer true of that object or situation. If something is green, and then is blue, this is a change, an event. If an emerald before our eyes were to turn blue after being green, we would seek an explanation of such event, some other change that caused the observed change. Our first stab at defining a change or event would appear to work. But wait: if an emerald were to stay green as time t ticked past, it would have changed from grue to bleen (something is bleen if observed before time t and is blue, and otherwise is green). The predicate “grue” would not have stayed true of it, for it would have come to instantiate the grue-excluding predicate “bleen”. Ducasse did not, as far as I know, suggest my “first stab” at defining a change or event, but neither did he volunteer any altemative.16 Without some idea of what constitutes a

l4 Ducasse’s most extended discussion of causality is in his Nature, Mind and Death, The

Is Ibid., p. 95. l6 Lawrence Lombard does seem to have accepted my ”first stab”. See his “Events and

Open Court Publishing Co., La Salle, Illinois, 1951.

Their Subjects”, Pacific PhiZosophicuZ QuarterZy 62 (1981), p.138.

272 Donald Davidson

change, his analysis of cause in terms of “only change’’ leaves us up in the air. Of course one thing that is green is similar to something else that is green; they are the same color. But this is no help. If grue and bleen aren’t colors, let us call them “tolors’’ - a property is a tolor if it changes color at time t. Then one thing that is grue is similar to something else that is grue; they are the same tolor.

There is an obvious correspondence between Hume’s problem and Du- casse’s problem. Hume’s problem was the problem of relevant generality; he needed to be able to say when one event was relevantly “similar” to another in order to distinguish lawlike generalizations from non-lawlike ones. Ducasse’s problem looked at first unrelated, since it apparently concerned particular cases; it arose in the course of trying to define cause in a way Ducasse thought was totally at odds with Hume’s approach, a way that makes no appeal to other cases. But in fact Ducasse’s definition of the causal relation has no content unless we are able to distinguish changes form non-changes,17 and this distinction turns out to involve generality in the sense that it is just the predicates which are projectible, the predicates or properties that enter into valid inductions, that determine what counts as a change.

The underlying problem is in both cases the same: neither Hume nor Du- caSse has specified when cases are relevantly similar. Hume needs to say when one change is relevantly similar to another (“Same cause, same effect”); Du- casse needs to be able to say when one state is not relevantly similar to another, i.e., when a change or event occurs. Thus Hume’s and Ducasse’s defini- tions of cause, so apparently odds, are essentially equivalent. Hume says that c caused eif and only if every event similar to cis followed by an event similar to e, that is, if and only if every event that is d ike is followed by an event that is elike. Ducasse says c caused e if and only if e is the only change that followed c, and cis the only change preceding e. Suppose the c-like event is the striking of a particular match and the elike event the lighting of that match. An event is clike if it is a striking of a match and elike if it is the lighting of a struck match. According to Hume, c caused e if and only if every clike event is followed by an elike event. According to Ducasse, c caused e if and only if c was the only change in a situation in which e was the only change that immediately followed. But cwas the cause of eonly if cand ewere the relevant changes, and

l7 Ducasse counts ”unchanges” as well as changes as events, but this terminological point is unrelated to the problem he faces in distinguishing events that are changes form events that are unchanges. In this essay, I do not follow Ducasse’s terminology: I call only changes events.

Laws and cause 273

they were this only if cwas a change from not being struck to being struck, i.e., a clike event, and e was a change from not being alight to being alight, i.e., was an elike event. This is the case, Ducasse argues, if and only if every event that is clike is followed by an event that is elike. These formulations are crude, but they serve to bring out the central fact that it is only if clike” and “elike” are the right sort of predicates, when taken together, that the quoted generalization-schemata are, if true, laws.

It is not surprising, then, that singular causal statements imply the existence of covering laws: events are changes that explain and require such expla- nations. This is not an empirical fact: nature doesn’t care what we call a change, so we decide what counts as a change on the basis of what we want to explain, and what we think available as an explanation. In deciding what counts as a change we also decide what generalizations to count as lawlike.

If the Big Bang left behind a uniformly expanding universe, we should expect that as we expand along with it, the intensity of the background micro- wave radiation will be the same in all directions. It is not; the radiation is measurably stronger in one direction than in others. This has been explained by assuming that the difference is due to the motion of our galaxy relative to the general expansion. Subsequent observations revealed, however, that over a vast area all neighboring galaxies are moving with us towards a common spot in the sky. This was in turn explained by the hypothesis of some immense but unobserved mass at that spot (“the great attractor”). In December of 1993, careful studies of the motion of galaxies four times further out in space than had previously been studied showed that they too were travelling in com- pany with the rest through the background radiation at about 500 kilometers a second. Nothing has turned up that isn’t moving with the crowd. So either the universe is uneven on a far larger scale than had been supposed or explained by present theory, or else the background radiation does not provide a true rest frame. In that case, says Princeton astrophysicist Bohdan Paczynski, there’s something wrong with our prior definition of what is at rest and what is moving. Exactly: if you can’t explain it using one assumption of what counts as a change, adopt new categories that allow a redefinition of change.

The history of physics is replete with examples of such adjustments in the choice of properties to define change, thus altering what calls for a causal explanation. At a certain level of common-sense physics, nothing is more static than a rock in the desert (unless, of course, someone picks it up, at which point the cause of its change of position is obvious, and a rough law surfaces). Even

As related by Faye Ham, “Galaxies Keep Going With the HOW”, Science, 259(1993), p. 31.

274 Donald Davidson

untutored observation may note a change in the temperature of the rock, but this is geared to an even more noticeable change in the relation of the sun to the rock. More serious science discovers changes in the positions of invisible particles in the rock which take place independent of changes in temperature; recognition of these changes allows for the causal explanation of a far larger range of phenomena.

One way science advances is by recognizing change where none was seen before. It can also work the other way around. Galileo sparked a revolution- ary improvement in physics when he proposed that uniform rectilinear motion not be treated as a change requiring an explanation, but as a steady state. The result was to give up the search for a cause of such motion and to treat only deviations from such motion as changes. Further advances made uniform rectilinear velocity only a special case of uniform acceleration; unchanges took over an even larger territory with the idea of not treating gravity as a force, so that the motion of a body along geodesics as defined by a spacial framework determined by the distributions of masses in space became a state not requiring a cause. l9

The dispositions to which we advert to explain what happens to objects, or the things they do, encapsulate the relation between causality and laws. We explain why the lump of sugar dissolved when placed in water by mentioning that it was water-soluble; something is water-soluble if placing it in water causes it to dissolve. We gain some understanding of why someone flew into a rage over a trifle if we know he was irascible; some one is irascible if small things cause him to be angry. The causal powers of physical objects are essential to determining what sorts of objects they are by defining what sorts of changes they can undergo while remaining the same object and what sorts of changes constitute their beginnings or ends. Our concept of a physical object is the concept of an object whose changes are governed by laws.

It would, as remarked before, be a large mistake to suppose that every way of refening to a cause or effect tells us how to characterize the change in terms suitable for incorporation in a law. “Hurricane Andrew” is a perfectly good phrase for picking out a particular event, but there is nothing in the concept of a hurricane that allows us to frame precise general laws about the causes or effects of hurricanes. If a hurricane is an event, it is a candidate for causal explanation; to say something caused it (the formation of a certain extreme low presswe area for example) is to claim that the changes involved can be described in terms that would serve to formulate a general causal law.

l9 My discussion here is based in part on Robert Cumrmns . , “States, Causes, and the Law of Inertia”, Philosophical Stuak, 29( 1976): 21-36.

Laws and cause 275

To revert now to the idea that what makes a statement lawlike is not a mat- ter of the projectible character of its individual predicates, but of the appropriate pairing or matching of predicates. Let us ask how projectible Goodman’s favorite projectible predicate “green” is. It picks out a class of objects we project early and easily. “AU emeralds are green” is lawlike, until, of course, de- feated by counterexamples. But even if not proven false, it is not a strict law; for when is something green? We are inclined to say something is green if and only if it looks green to normal observers under normal conditions. But “looks green” and “normal” are not predicates that can be sharply defined, and they certainly cannot be reduced to the predicates of physics. They cannot feature in the laws of a closed system. If we imagine that a satisfactory definition of “emerald” can be devised in the vocabulary of physics, then “green” is not strictly projectible of emeralds. Of course, “green” is more projectible of emeralds than “grue”; but “green” and “blue” are only as projectible of emeralds and sapphires as “grue” and “bleed’ are of emerires and sapphalds. This distinction, between predicates like “green” and “grue” on the one hand, and predicates that together can feature in strict laws, is crucial, as I said, to the argument for Anomalous Monism. It is what distinguishes psychophysical laws from the laws of an advanced physics.

We are born, as Quine has emphasized, *O treating some pairs of things as more similar than others. We react differentially to sudden loud noises, and since we do not like such sounds, we soon learn to cringe from what has fre- quently preceded them. Thus long prior to the acquisition of language, or anything that can properly be considered concept formation, we act as if we had learned crude laws. We are inducers from birth; if we were not, infant mortality would be the rule, if there were any infants. Concepts, concep- tualized laws, the idea of causal relations between events, build on these foun- dations. In the course of avoiding and seeking, learning to control our envi- ronment, failing and succeeding, we build the lawlike habits that promote sur- vival and enhance life. These laws of action are highly pragmatic not only in their conspicuous ties to action, but in their breezy disregard of the irrelevant or implausible. The generalizations on which not only the untutored infant but also our adult selves mostly depend are geared to the “normal”, the “usual”. We don’t know, and for practical purposes don’t care, what would happen if there were no oxygen, temperatures were to fall to absolute zero, or there were a black hole in the closet (I now learn that there may be!). But practical purposes can change, and the conditions at one time happily abandoned

In ‘‘Natural Kinds“, in Ontological Relativity and Other Essays, Columbia University Press, 1969.

276 Donald Davidson

to ceteris paribus clauses may become relevant, or come to be recognized as relevant. We then refine our classifications to improve our laws.

Learning laws of greater generality provides no reason for jettisoning our proviso-laden work-a-day causal laws. When I flick on the light switch, I have no reason to reflect on the speed of light, though I may briefly dwell on that magic constant when my telephone calls to Perth are relayed - and percep- tibly delayed - by satellite technology. The more precise and general laws are, the less likely it is that we will be in a position to employ them in predicting the outcomes of our ordinary actions or the weather. Our intense interest in the explanation and understanding of intentional behavior commits us irrevoc- ably to such concepts as belief, desire, intention and action; yet these are concepts that cannot, without losing the explanatory the power they have which binds us to them, be reduced to the concepts of an all-encompassing physics.

We have interests that are not practical. There are things we want to understand whether or not we can control them and whether or not such knowledge will serve our mundane needs. Pursuit of the truth in such cases can in principle proceed without the constraints of practical control and gain. In this mood we can seek laws that have no exceptions.

The same strategy that serves to refine our practical lore, the strategy that leads us to adjust what counts as a change or as requiring a causal explanation, works here too, as is evident in the advances in physics due to Galileo, Newton and Einstein.

Robert Cummins puts it this way:

A decision about how to characterize [a statel- a choice of state variables - imposes distinctions between states and non-states, and hence determines what is and what is not construed as an effect. Such decisions are not arbitrary, in part because they have this consequence. Effects require direct causal explanation; if there is none to be had which statisfies, then we shall alter our taxonomy. This sounds like metaphysics, and indeed it is. . . All this presupposes a certain explanatory strategy [which] is easy to state in outline: what requires explanation is change, and changes are to be explained as effects, the trick being to characterize matters in a way which makes this possible, i.e., in a way which distinguishes genuine changes from states. 21

Cummins notes that physics leading up to and including classical mechanics “and its kin” has followed the strategy he describes and adds, somewhat glumly, “this strategy seems to be breaking down at the quantum level”. I am not up to evaluating every implication of this last suggestion, but I do not think quantum physics poses a threat to the cause-law thesis. If what Cum- mins means is that the ideal of a completely deterministic theory must be

21 bid., p. 33.

Laws and cause 277

given up if quantum physics is the last word, he is of course right. But his strategy would seem to apply to quantum physics if we delete the word “direct” from the requirement that effects require direct causal explanation, and understand the changes that are to be explained as sometimes being changes in probabilities.

This is the view promoted many years ago by Henry Margenau. He formu- lates what he calls the “Principle of Causality” as follows: “causality is viol- ated when a given state A is not always followed by that same state B”. 22 Like Cummins, he notes that without an understanding of what constitutes a state, this tells us nothing; but that, on the other hand, once we decide what constitutes a state, we have decided what counts as a causal law. Why, though, should we count on the there being a way of specifying states that allows the formulation of exceptionless laws? Here is Margenau’s answer:

[Wlhenever a physical system does not appear to be closed, that is when the dif- ferential equations describing it contain the time explicitly. . ., we conclude that the variables determining the state in question are not completely known. We then look immediately for hidden properties whose variation may have produced the inconsisten- cies, and whose inclusion in the analysis would eliminate them; moreover if we do not find any we invent them. This procedure is possible because. . . the term “state” is un- defined. . . it seems, then, that the causality postulate reduces to a definition of what is meant by “state”. 23

For this reason, Margenau says, “physics can never inform us of a failure of the causality principle”. Quantum mechanics, he concludes, poses no threat to the causality principle; it just once again redefines the concept of a state.

Margenau could not, of course, have known about Bell’s inequality theo- rem, nor the experiments that made use of it to prove that there can be no local hidden variable theory. But Margenau did not understand the Principle of Causation to require a deterministic physics: he required only that the notion of a state be formulated in such a way as to insure that the laws be strict. The point comes out clearly in a recent discussion of the philosophical fallout from the experimental proof of the impossibility of a hidden variable theory:

Complete knowledge of the state of a classical system at a given time is synonymous with exact knowledge of all observables at that time. Complete knowledge of the state of a quantum mechanical system is ensured by the exact knowledge of only a subset of all observables at the same time. . ’. Only when one asks for the value of an observable not in the complete set of observables that are knowable simultaneously does one en- counter the probabilistic nature of QM. In this sense the probabilistic element enters

22 “Meaning and Scientific Status of Causality”, Philosophy of Science, 1 (1934), p. 140. 23 hid., pp. 144, 5.

278 Donald Davidson

QM through the measurement process, which is bound to change the observed system. . . The various possible outcomes of such a measurement occur with probabilities that can be predicted exactly. The outcome of any one such measurement. . . cannot be predicted.

Thus Margenau is borne out: the concept of a state of a quantum mechanical system has been defined to ensure that all that can be known about it at a moment completely determines the state. The identity of a quantum state is sensitive, in a way that states in classical systems are not, to the effect of measurements. In a classical system, it is assumed that the effect of measurements can in principle be reduced to an arbitrarily high degree, while in quantum physics the effect of measurements is integral to the theory: the measur- ing device becomes part of the same physical system as what is measured, and subject to the same laws. Quantum mechanics sacrifices determinism as the cost of gaining universality. “It states beautifully what is completely ignored in classical physics: through man nature can observe itself”. 25 Far from challeng- ing the cause-law thesis, quantum physics exemplifies it.

v. Quantum physics may not, of course, be the last word. It is not only Ein-

stein who dreamed of a theory that would supersede quantum theory, or in effect make quantum theory deterministic. Steven Weinberg, for example, not only dreams of a unified Theory of Everything, but speculates that such a theory might be completely deterministic. 26 But it is hard to think that the question whether such a theory exists is a purely conceptual question, at least a theory human beings could, even in principle, invent and test. Surely we must allow that the best physics that is possible for us is irreducibly probabilistic. Does this mean the cause-law thesis is not tenable? I think not. Margenau’s Principle, essentially Hume’s “same cause, same effect” principle, put us on the track of the idea that natural and devised standards of similarity play a fundamental role both in the notion of change and the notion of law. This pro- vided a legitimate connection between the concept of causality and the concept of a law. We can imagine this connection being tight enough to support completely deterministic laws. Indeed, this is how classical mechanics assumed things were. But if physics cannot be made deterministic, if the ulti-

Fritz Rohrlich, “Facing Quantum Mechanical Reality”, Science, 221 (1983), p. 1254. 25 Ibid., p. 1253. 26 In Dreanzr of a Final Theory, Hutchinson, as reported by John Leslie, The Times Llte-

mry Supplement, Jan. 29, 1993, p. 3.

Laws and cause 279

mate laws of the universe, so far as we will ever know, are probabilistic, then we must think of causality as probabilistic. Singular causal statements will still entail the existence of strict laws, even at the quantum level, but the laws will not meet Hume’s or Kant’s or Einstein’s standards.

Dialectics Vol. 49, No 2-4 (1995)

laws and cause

Documents